[OSRM-talk] Routing across features with "issues"

[Andrew King]

Hi Richard,

Thank you for your guidance. Yes, I can see that a lot of effort is required to get the datasets working cleanly. We have been able to make some adjustments resulting in improvements which are in better keeping with the requirements of our customers. But changing the rules and configuration can only take you so far. Errors in the data will always be problematic.

Thank you for your insight. We will bear this in mind as we move forward.

Kind regards
Andrew


________________________________
From: Richard Fairhurst [richard at systemed.net]
Sent: 08 June 2021 19:14
To: osrm-talk at openstreetmap.org
Subject: Re: [OSRM-talk] Routing across features with "issues"

Andrew King wrote:
> What is the best way to deal with issues such as this? I could obviously
> fix the data for this one case, but there are probably thousands of
> similar cases such as this.

Julien has explained the issue in this particular case ("dupe" nodes).

As a more general word of warning:

If you want good results from pedestrian (or bike) routing with OSM, no matter what router you use, you will have to put in a lot of work. It isn't an out-of-the-box solution. OSM path and access data is messy and inconsistent, particularly across different countries. You will have to do a lot of data fixup and a fair amount of coding (in OSRM's case, in the Lua profiles) to get something good - my profiles and other pre-processing code for cycle.travel are well into the thousands of lines of code.

As an example of the sort of issues you can encounter:
 https://www.openstreetmap.org/changeset/105985206

Here you can see an angry landowner who is finding people are being diverted across his land by routing apps. I'm pretty sure I know what's happening here - certain routing apps, perhaps in this case AllTrails and Komoot, are not very good at understanding how tagging conventions differ between countries. And bear in mind that's with a staff of over 60 (Komoot) or $75m of VC (AllTrails).

I don't want to put you off - pedestrian/bike routing with OSM, when done right, is much better than with any other dataset. And, of course, it's up to you to decide the return on satisfied routing customers vs hours invested in tuning the routing. But if you want to get it right, it takes a lot of time.

cheers
Richard
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/osrm-talk/attachments/20210608/95c34c8f/attachment-0001.htm>